From AI Experimentation to Measurable Outcomes: What I’ve Learned Building (and Using) AI in HR

I spend a lot of time talking to other CHROs right now, and the conversation is almost always the same.

Everyone has more data than they know what to do with. Everyone is experimenting with AI. And almost no one feels like it’s actually changing behavior at scale.

I say that with empathy because not long ago, we were in the same place.

What makes my role a little different is that I’m not just buying HR technology, I’m also helping build it. My team at Perceptyx is often the first to use the products we bring to market. That’s a gift, but it’s also very exposing. When something doesn’t work, there’s nowhere to hide.

And for a while, we were running into the same wall I hear from everyone else: We had the insights. We just weren’t seeing enough change.

The Problem We Had to Admit

We’ve been in the listening space for a long time. We’re very good at understanding what employees are experiencing and where things are breaking down.

But knowing isn’t the same as improving. We could tell you where recognition was low. We could show where engagement was slipping. We could highlight gaps in manager effectiveness.

What we couldn’t guarantee was that anything would actually happen next.

Our managers were busy. Employees were stretched. And like most organizations, we had built systems that required people to step outside of their day-to-day work to take action.

That’s where things can easily fall apart. We uncovered a stat that really forced the issue for us:

46.2% of our employees said the biggest barrier to achieving their career goals was lack of time or capacity.

Not lack of interest. Not lack of resources. Time!

At that point, I realized that providing more insights or adding more programs wasn’t going to solve anything. We had to change how our employees were learning and evolving and make sure it was happening directly inside the business in the flow of work.

What Actually Started to Work

The biggest shift for us came when we started focusing less on insight delivery and more on what happens in the moment a manager can do something about it.

When Perceptyx rolled out our AI-native activation agent, I was the first to lean in. Being in my position, I was feeling the urgency to start activating all of our data just as much as our customers were.

Instead of expecting managers to go into a dashboard, interpret data, and decide what to do, we were now bringing the action to them in their natural workspace.

We started delivering targeted, AI-driven nudges and coaching based on real team feedback. Not theoretical guidance. Not leadership platitudes.

Real prompts, tied to what was happening on their team. For example, a manager might get a nudge that recognition on their team has dropped, along with a suggestion to acknowledge specific recent contributions. Or a prompt to follow up with a team member who hasn’t had a recent check-in. The difference is subtle, but it matters. We weren’t asking managers to spend time analyzing more data. We were helping them act immediately in the flow of their work.

And when they act consistently, things change.

We found that our teams with managers who regularly engaged with these nudges saw a 5-point increase in engagement year over year. That was a turning point for me.

For the first time, I felt confident saying we weren’t just collecting feedback anymore—we were influencing behavior in a real and measurable way. If behavior changes and capability improves, execution improves. That’s the connection we’re validating internally. Every reduction in preventable attrition, every manager who becomes measurably more effective — that compounds directly into our ability to execute.

What Didn’t Work (and What I See Others Getting Wrong)

We didn’t get here cleanly. Early on, we made the same mistake I see everywhere. We treated AI like an add-on.

We layered it into existing processes and expected different results. We generated more insights without making it easier to act on them. In some cases, we actually made things more complicated for managers.

That doesn’t scale.

Another lesson: generic doesn’t work.

If a nudge could apply to anyone, it’s easy to ignore. The only things that consistently drove action were specific, timely, and clearly connected to the manager’s team.

And I’ll say this plainly, because I think it gets glossed over: Manager behavior is the lever.

If managers aren’t changing what they do day to day, nothing else matters. You can have the best insights, the best tools, the best intentions, but it won’t show up in outcomes.

Catch more HRTech Insights: HRTech Interview With Hari Kolam, CEO and Co-founder of Findem: Featuring Findem’s GliderAI

Where This Leaves Us on the Skills Gap

Turning insight into action solved one part of the problem for us.

But it also made something else very clear. Prompting behavior is very powerful but building new employee capability is harder.

We can nudge a manager to give better feedback. But if they don’t actually know how to coach, how to develop people, how to have difficult conversations, there’s a ceiling to the impact.

At the same time, the broader challenge hasn’t gone away. The skills gap is still widening, and most organizations are trying to close it with approaches that don’t match the pace of change.

This is where we’re evolving next.

What We’re Working Toward Now

Recently, we acquired a company called Lyceum and have added their AI-native employee development product to the Perceptyx platform. Lyceum, now Develop, is a conversational learning agent that adapts to individual users while verifying comprehension and knowledge gained.

One thing I know for sure is that leaders often leave workshops inspired but revert within days because work conditions haven’t changed. That’s why we’re focused on changing behavior in the moment of work—not reminders, not a class at 9:00 a.m. on a Wednesday, but guidance at the point of decision.

This is new for us, just as much as it is for our customers, and my team is actively integrating it into our processes. I’m learning alongside them so we can optimize how our systems work together and create a continuous loop from insight to action to development—ensuring we’re not just identifying gaps, but actively building the capabilities our people need and reinforcing them over time.

Our vision is very intentional: connecting what we hear from employees, what we prompt managers to do, and how employees build skills into one continuous loop.

In practice, that means using feedback to identify where skills need to be developed, delivering learning in small, usable moments—not long-form training—and reinforcing that learning over time through the same nudges that are already driving behavior.

Our goal is to make development part of the work itself, not something people have to carve out extra time for. When a manager is about to walk into a difficult conversation, or when a team is showing signals of change fatigue, the system surfaces the right action at that moment.

Because if time is the biggest barrier, we have to design around that reality. This is how we begin to close the skill gaps we’re seeing across our organizations—by connecting fragmented systems and leveraging AI to create a true loop from insight to action to development, and ultimately, transformation.

What I’d Tell Other HR Leaders Right Now

Most organizations don’t have an insight problem anymore! They have an action problem. And increasingly, a capability problem.

If I were starting from scratch, I wouldn’t begin with tools. I would start with a few questions:

• Where do we actually need behavior to change?
• When does that behavior happen in the day-to-day flow of work?
• How do we support it in-the-moment?

I would be honest about how we measure success. A lot of what we’ve historically tracked in HR tells us what we launched, not what changed.

That gap matters because AI isn’t going to differentiate you on its own. Plenty of organizations will have access to the same capabilities.

What will matter is whether you can take insight and turn it into consistent action and then build the skills to sustain it. That’s the work we’re in the middle of now.

We know our environment changed faster than our management practices.

We know AI is reshaping roles monthly. Capability decays faster than annual planning cycles. Employees today don’t want another engagement initiative, and we should take that as a signal that the organization can handle what’s coming.

That means leadership behavior has to adjust in real time — not after a training session or an engagement survey 8 months later.

The organizations winning right now have figured out how to sense, act, and learn faster. That’s the shift we’re building toward.

And if I’m being honest, it’s the first time in a while that I feel like we’re not just understanding our people better, we’re actually helping them work differently. From insight to action to development — and ultimately, measurable behavior change!

About Perceptyx

Perceptyx is the People Activation company that unites employee listening and people development in one AI-powered system — signal, action, and skill running on a single native data model — so organizations can discover what needs to change, activate the behaviors that drive it, develop the capabilities to sustain it, and prove the impact.

Read More on Hrtech : Why SWIFT is Too Slow for Your Global Workforce?

[To share your insights with us, please write to psen@itechseries.com ]